Where Tournament Pressure Breaks Cricket's Baseline: Dot-Ball Weight, Expected Wickets, and the Death-Over Ledger

One match from the latest tournament is still marked in red ink in my notebook. A target of 180. After four overs, 38 for three. The scoreboard calls that a collapse; my table calls it something else. At that venue, the average return per ball after the 16th over was 1.42. The required rate stood at 1.48 — exactly zero point zero six above baseline. My win-probability model read 41 percent. The commentary box was saying the match was almost gone. By the 27th over, at 149 for four, the model had climbed to 63 percent. Twenty-four balls left, 31 needed. The side won with 14 balls to spare. By next morning, social feeds had turned it into an impossible comeback.

I have watched cricket for fourteen years and counted it for eight. Sitting in the stands, I do not look at the scoreboard; I look at the venue baseline. The first model I built alone did not predict football — it predicted my own patience. Cricket enforces that lesson harder, because every ball is a discrete event with a price.

Context: How the Baseline Is Built

My phase baseline rests on four inputs: ball-by-ball sequence, venue code, innings phase, and the opposition bowling profile. On every delivery I log three numbers — expected runs, expected wickets, and dot-ball pressure. The first comes from line, length, bounce, stroke zone, and field setting. The second uses the same features, with the outcome variable being dismissal, batter-bowler matchup, and over position.

Dot-ball pressure I keep as the ratio of dot balls to wicket chances. It does not map directly onto a football pressing metric, because cricket pressure is measured in ball-by-ball rhythm. Still, it is the single index that does most work in my editorial tables, because it measures bowling dominance beyond the scoreboard.

A question I get regularly: should the baseline be match-specific? My answer is no. It should be venue- and phase-specific, not match-specific. A match-specific baseline means explaining the past while knowing the result. That is a story, not a model.

In my first two seasons I built baselines only from top-tier team data. That was wrong. Against weaker opponents almost any bowling plan works, and that inflates the baseline artificially. I now use league-level weighting and print the sample size in every report. I also print the data provenance question, because feeds from Dhaka and Manchester can count different dot balls in the same match.

Core: The Six-Match Deviation Chain

Across the tournament's first six matches, the powerplay average is 50.2 against a baseline of 47.8. The deviation is plus 2.2, inside one standard error. Powerplay aggression has risen — that is normal variance, not a story. Openers like Jos Buttler are striking at 1.86 runs per ball against a baseline of 1.61, but on a six-match sample that is not yet a signal.

The picture changes in the middle overs, seven through fifteen. The tournament average is 67.1 against a baseline of 74.3 — a deviation of minus 7.2, with a standard error of 1.9. This is the tournament's first real deviation: middle-over runs have fallen because spinners are over-pressing and double-paced pitches are producing more dot balls. Spinners in the Rashid Khan and Shakib Al Hasan mould have dropped to 1.02 runs per ball in this phase, where the tournament baseline is 1.28. When an all-rounder like Hardik Pandya bowls in the middle, spin pressure and pace pressure merge, and the run rate falls hardest there.

In the death overs the ledger flips again. From the sixteenth to the twentieth, the tournament average is 52.4 against a baseline of 48.9 — plus 3.5. In the same window dot-ball rate has risen from 27 to 31 percent. Dot balls are rising and strike rate is rising together. When those two move in tandem, you get what I call binary death: boundary or dot, nothing between. Roughly 41 percent of boundaries have come on the leg side, because wide-yorker usage is down nine percent.

Expected wickets across the tournament total 43.6, while 48 wickets have actually fallen — a deviation of plus 4.4. Batters have given away more wickets than expected. Jasprit Bumrah's death-over economy is 7.1 against a baseline of 9.4; that single number explains India's death-over defence, and every other explanation written around him is a story.

In the dot-ball pressure table, spin accounts for 61 percent of tournament pressure and pace 39. In the death overs the ratio inverts: pace 74, spin 26. That phase switch is the tournament's most repeated pattern. Yet in five matches at least one spinner has bowled in the last two overs, and those overs produced the costliest returns. A side that keeps a spinner for the death overs is paying 2.1 more expected runs per over.

Fielding residual is hard to isolate because a dropped catch is a rare event. I still keep a diving-catch conversion column. Six matches produced nine diving chances and six catches. The baseline conversion rate in that zone is 38 percent. Fielding sides have beaten the baseline, and that surplus explains most of the expected-wicket deviation.

Case Study: The 68-Run Powerplay That Lost

In one tournament match a side made 68 in the powerplay against a baseline of 47.8 — a deviation of plus 20.2. They still lost. From seven to fifteen they made 52 against a baseline of 74.3, and lost five wickets there. Expected wickets said 2.4 should have fallen. The fielding side took three catches whose baseline conversion probability was 38 percent.

Germany did not lose to South Korea; they lost to 26 shots and no goals. The cricket equivalent is: the side did not lose in the powerplay, it lost in the middle overs to 52 runs. The lesson from Kazan in 2026 — 74 percent possession, 26 shots, 2.7 xG — holds. Territory and dominance are two different things.

Contrarian: Correlation Is Not Causation

On a six-match sample, an easy story forms: attack with spin in the middle overs and you win. Correlation here is not causation. Sides that keep spinners in the middle overs are usually the side behind, so they set aggressive fields. The spin-usage and result relationship is therefore partly generated from the other direction.

I ran a placebo test: splitting the tournament data randomly into two halves, the spin-over-ratio and win relationship disappeared in half the runs. With six matches, a causal claim cannot survive a 95 percent confidence interval. What can be claimed is direction: sides leaving spin for the death overs are conceding 2.1 more expected runs per over, and that direction held in four matches.

I do not chase narratives; I build a table and wait for them to arrive. The eye test is a witness; the data is the cross-examination. In tournament talk the word momentum returns every night without anyone offering an operational definition. I tried one: runs per ball in the three overs after two consecutive wickets. Across six matches the number is 1.44, against a normal rate of 1.51. Momentum is not visible here.

Auditing the Baseline Itself

Treating the baseline as sacred is its own danger. Pitches used in this tournament differ from last season, two new balls are in play, and the day-night dew factor has shifted. A baseline built on the previous three seasons is slightly stale against that reality. So every report now carries an era-adjusted column beside the baseline.

Labelling differences between Bangladeshi and UK feeds are not small either. One feed merges out-swing and seam movement into a single variable; the other keeps them apart. In one match last year the dot-ball gap between the two feeds was four percent. However good the model, a dirty pipeline produces a dirty decision. On a 20,000-ball dataset, four percent is 800 balls — not a small gap.

Home Advantage and the Empty-Stadium Lesson

In 2026 I counted the silence and found it had a home advantage. Across the first five Bundesliga rounds, home win rate fell from 43.2 to 21.1 percent, and home goals per game from 1.65 to 1.08. Every empty stadium was a controlled experiment nobody asked for.

In a cricket tournament I never get that control fully, only partially. I am tracking how the death-over pressure index moves when crowd ratio shifts at neutral venues. The early read: a ten percent rise in crowd density cuts the home-near side's death-over economy by 0.2. The number is small, but the direction is clean, and for now that is enough.

Takeaway: What I Watch Next Round

Three things next round. Whether the spin-over ratio in the middle overs climbs above the tournament average of 42 percent, and the dot-ball rate with it. Whether high powerplay scores are forming a negative relationship with middle-over runs, meaning aggression is shifting from one phase to another. And whether death-over yorker usage returns to its earlier level, because the wide-yorker deficit is the tournament's biggest tactical gap right now.

I do not know who lifts the trophy. I know which numbers will change next week and which will not. That much is the job. Build a table, wait, and when the story arrives, ask only one question — what is your sample size?